华南大学研究人员王和吴发表论文，宣称在OpenAI Codex的协助下证明了球面Hadwiger猜想1。这一猜想自1974年以来已保持开放状态长达四十余年1。根据论文的免责声明，OpenAI Codex被用于"开发证明细节、识别空缺、组织排版和编辑英文"1。
这一进展引发了对AI在数学研究中角色的深层思考。OpenAI在GPT-4发布报告中估计，数学家工作对大型语言模型的暴露度达到100%，这一比例高于作家、翻译、艺术家等其他创意工作1。研究者认为，数学研究的经济价值相对较低、训练数据量也相对有限，这导致AI在该领域的应用出现较晚1。同时，这一发展也引发了关于数学研究职业前景的担忧，包括学术激励体系可能面临的根本性改革，以及研究数学可能面临类似古典音乐的地位转变1。
Researchers from South China University announced the use of an AI assistant to help resolve a longstanding mathematical problem, raising fundamental questions about the future role of artificial intelligence in mathematical research. Wang and Wu published a paper describing how they employed OpenAI Codex to assist in proving the spherical Hadwiger conjecture, which has remained open for over forty years since 1974.1
According to a disclaimer in their paper, OpenAI Codex was utilized for "developing proof details, identifying gaps, organizing formatting and editing English text."1 The development underscores broader concerns about how AI capabilities may reshape mathematical practice. OpenAI's GPT-4 technical report estimated that mathematician work faces 100% exposure to large language models—a rate exceeding that of writers, translators, and artists among other creative professions.1
Observers have drawn parallels to the professionalization and institutional protection of classical music, suggesting that mathematical research itself may face significant transformation.1 The underlying factors contributing to this shift include the relatively modest economic value of mathematical research and smaller volumes of training data available in mathematics compared to other fields, which has historically slowed AI's penetration into the discipline.1 The developments have prompted discussion about whether fundamental reforms to academic incentive systems may become necessary to address the evolving landscape of mathematical discovery and validation.1
评论
还没有评论，欢迎留下第一条。